[gmx-developers] Reference coordinates in mdrun - Availability at all nodes
Berk Hess
hess at cbr.su.se
Wed Jan 13 11:35:11 CET 2010
Hi,
I just have a suggestion for the name:
group_util is a bit too general, I would suggest something like groupcoord.
Berk
Carsten Kutzner wrote:
> Hi,
>
> as David suggested, I have assembled in a new file the routines that are
> useful to do comparisons to a set of reference positions (and for other
> things as well). I called the new files group_util.[c,h] and put them
> in ./src/mdlib/. I attach the header file if someone wants to take a look
> over it and has useful suggestions before I commit. At the moment, the routines
> are used by the essential dynamics module, and most of them also by the
> enforced rotation module (which is in a separate branch, though).
>
> Tsjerk, you might want to use dd_make_local_group_indices() to select
> the local atoms of the group you need to compare against the reference.
> If you pass the the cross products of particle velocities and reference
> positions per node to get_center_comm(), it will do the global summing for
> you.
>
> If everybody is happy with the new file, I would commit it to the master
> branch.
>
> Carsten
>
>
> David van der Spoel wrote:
>
>> Carsten Kutzner wrote:
>>
>>> On Jan 6, 2010, at 12:27 PM, David van der Spoel wrote:
>>>
>>>
>>>> Tsjerk Wassenaar wrote:
>>>>
>>>>> Hi Carsten,
>>>>> Thanks a lot, that seems quite useful, and comes very close to what I
>>>>> need. To elaborate a bit more, at runtime, I only need to have
>>>>> differences between particle positions and their respective reference
>>>>> positions, so adaptive local indexing seems most efficient, even
>>>>> though it requires more memory. That shouldn't be a big problem
>>>>> though. The difference vectors are to be summed, but I'd only need to
>>>>> communicate partial sums, and then redistribute the outcome. I'm
>>>>> vigilant, as I don't want to compromise/break anything. Would you mind
>>>>> to look a bit over my shoulder? I can send you a more full account of
>>>>> what I'm doing off line...
>>>>>
>>>> If the simlarity between the code is sufficient, I would like to
>>>> advocate to take out the communication bits from edsam.c and move
>>>> them to a separate source code file, used by edsam and Tsjerk's code.
>>>> In general the code that does communication is most prone to problems
>>>> when changing e.g. the parallellisation scheme, and hence it is
>>>> useful to minimize the amount of code doing communication. IMHO it is
>>>> even worth doing some unnecessary work (e.g. copying arrays in
>>>> running mdrun) in order to be able to re-use communication code.
>>>>
>>>>
>>> This is a good idea. In general, all approaches that make use of some
>>> set of reference positions against which the current ones are compared
>>> could use these routines (this would also apply to the enforced rotation
>>> module). Maybe we could also include other useful operations on
>>> "collective"
>>> coordinate subsets like fitting to the reference or getting the center
>>> of mass, or making the structure whole.
>>>
>> That sounds good!
>>
>> After moving the relevant routines from edsam.c to a new file it would
>> be good to make a small API (in the header file) with comprehensible
>> names not related to edsam.
>>
>> Cheers,
>>
>
>
> /* -*- mode: c; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4; c-file-style: "stroustrup"; -*-
> *
> *
> * This source code is part of
> *
> * G R O M A C S
> *
> * GROningen MAchine for Chemical Simulations
> *
> * Written by David van der Spoel, Erik Lindahl, Berk Hess, and others.
> * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
> * Copyright (c) 2001-2008, The GROMACS development team,
> * check out http://www.gromacs.org for more information.
>
> * This program is free software; you can redistribute it and/or
> * modify it under the terms of the GNU General Public License
> * as published by the Free Software Foundation; either version 2
> * of the License, or (at your option) any later version.
> *
> * If you want to redistribute modifications, please consider that
> * scientific software is very special. Version control is crucial -
> * bugs must be traceable. We will be happy to consider code for
> * inclusion in the official distribution, but derived work must not
> * be called official GROMACS. Details are found in the README & COPYING
> * files - if they are missing, get the official version at www.gromacs.org.
> *
> * To help us fund GROMACS development, we humbly ask that you cite
> * the papers on the package - you can find them in the top README file.
> *
> * For more info, check our website at http://www.gromacs.org
> *
> * And Hey:
> * Gallium Rubidium Oxygen Manganese Argon Carbon Silicon
> */
>
> /*! \file group_util.h
> *
> * @brief Assemble atom positions for comparison with a reference set.
> *
> * This file contains functions to assemble the positions of a subset of the
> * atoms and to do operations on it like determining the center of mass, or
> * doing translations and rotations. These functions are useful when
> * a subset of the positions needs to be compared to some set of reference
> * positions, as e.g. done for essential dynamics.
> *
> */
>
> #ifdef HAVE_CONFIG_H
> #include <config.h>
> #endif
>
> #include <stdio.h>
> #include "typedefs.h"
>
>
> /*! \brief Select local atoms of a group.
> *
> * Selects the indices of local atoms of a group and stores them in anrs_loc[0..nr_loc].
> * If you need the positions of the group's atoms on all nodes, provide a coll_ind[0..nr]
> * array and pass it on to communicate_group_positions. Thus the collective array
> * will always have the same atom order (ascending indices).
> *
> * \param ga2la[in] Global to local atom index conversion data.
> * \param md[in] Pointer to runtime atom data.
> * \param nr[in] The total number of atoms that the group contains.
> * \param anrs[in] The global atom number of the group's atoms.
> * \param nr_loc[out] The number of group atoms present on the local node.
> * \param anrs_loc[out] The local atom numbers of the group.
> * \param nalloc_loc[inout] Local allocation size of anrs_loc array.
> * \param coll_ind[opt] If not NULL this array must be of size nr. It stores
> * for each local atom where it belongs in the global
> * (collective) array such that it can be gmx_summed
> * in the communicate_group_positions routine.
> */
> extern void dd_make_local_group_indices(gmx_ga2la_t ga2la, t_mdatoms *md,
> const int nr, int anrs[], int *nr_loc,
> int *anrs_loc[], int *nalloc_loc,
> int coll_ind[]);
>
>
> /*! \brief Assemble local positions into a collective array present on all nodes.
> *
> * Communicate the positions of the group's atoms such that every node has all of
> * them. Unless running on huge number of cores, this is not a big performance impact
> * as long as the collective subset [0..nr] is kept small. The atom indices are
> * retrieved from anrs_loc[0..nr_loc]. If you call the routine for the serial case,
> * provide an array coll_ind[i] = i for i in 1..nr.
> *
> * \param cr[in] Pointer to MPI communication data.
> * \param xcoll[out] Collective array of positions, idential on all nodes
> * after this routine has been called.
> * \param shifts[inout] Collective array of shifts for xcoll, needed to make
> * the group whole. This array remembers the shifts
> * since the start of the simulation (where the group
> * is whole) and must therefore not be changed outside
> * of this routine!
> * \param extra_shifts[buf] Extra shifts since last time step, only needed as
> * buffer variable [0..nr].
> * \param bNS[in] Neighborsearching/domain redecomposition has been
> * performed at the begin of this time step such that
> * the shifts have changed and need to be updated.
> * \param x_loc[in] Pointer to the local atom positions this node has.
> * \param nr[in] Total number of atoms in the group.
> * \param nr_loc[in] Number of group atoms on the local node.
> * \param anrs_loc[in] Array of the local atom indices.
> * \param coll_ind[in] This array of size nr stores for each local atom where
> * it belongs in the collective array so that the local
> * contributions can be gmx_summed. It is provided by
> * dd_make_local_group_indices.
> * \param xcoll_old[inout] Positions from the last time step, used to make the
> * group whole.
> * \param box[in] Simulation box matrix, needed to shift xcoll such that
> * the group becomes whole.
> */
> extern void communicate_group_positions(t_commrec *cr, rvec *xcoll, ivec *shifts,
> ivec *extra_shifts, const bool bNS,
> rvec *x_loc, const int nr, const int nr_loc,
> int *anrs_loc, int *coll_ind, rvec *xcoll_old,
> matrix box);
>
>
> /*! \brief Calculates the center of the positions x locally.
> *
> * Calculates the center of mass (if masses are given in the weight array) or
> * the geometrical center (if NULL is passed as weight).
> *
> * \param x[in] Positions.
> * \param weight[in] Can be NULL or an array of weights. If masses are
> * given as weights, the COM is calculated.
> * \param nr[in] Number of positions and weights if present.
> * \param center[out] The (weighted) center of the positions.
> *
> */
> extern void get_center(rvec x[], real weight[], const int nr, rvec center);
>
>
> /*! \brief Calculates the sum of the positions x locally.
> *
> * Calculates the (weighted) sum of position vectors and returns the sum of
> * weights, which is needed when local contributions shall be summed to a
> * global weighted center.
> *
> * \param x[in] Array of positions.
> * \param weight[in] Can be NULL or an array of weights.
> * \param nr[in] Number of positions and weights if present.
> * \param dsumvec[out] The (weighted) sum of the positions.
> * \return Sum of weights.
> *
> */
> extern double get_sum_of_positions(rvec x[], real weight[], const int nr, dvec dsumvec);
>
>
> /*! \brief Calculates the global center of all local arrays x.
> *
> * Get the center from local positions [0..nr_loc], this involves communication.
> * Not that the positions must already have the correct PBC representation. Use
> * this routine if no collective coordinates are assembled from which the center
> * could be calculated without communication.
> *
> * \param cr[in] Pointer to MPI communication data.
> * \param x_loc[in] Array of local positions [0..nr_loc].
> * \param weight_loc[in] Array of local weights, these are the masses if the
> * center of mass is to be calculated.
> * \param nr_loc[in] The number of positions on the local node.
> * \param nr_group[in] The number of positions in the whole group. Since
> * this is known anyway, we do not need to communicate
> * and sum nr_loc if we pass it over.
> * \param center[out] The (weighted) center of all x_loc from all the
> * nodes.
> */
> extern void get_center_comm(t_commrec *cr, rvec x_loc[], real weight_loc[],
> int nr_loc, int nr_group, rvec center);
>
>
> /*! \brief Translate positions.
> *
> * Add a translation vector to the positions x.
> *
> * \param x[inout] Array of positions.
> * \param nr[in] Number of entries in the position array.
> * \param transvec[in] Translation vector to be added to all positions.
> *
> */
> extern void translate_x(rvec x[], const int nr, const rvec transvec);
>
>
> /*! \brief Rotate positions.
> *
> * Rotate the positions with the rotation matrix.
> *
> * \param x[inout] Array of positions.
> * \param nr[in] Number of entries in the position array.
> * \param rmat[in] Rotation matrix to operate on all positions.
> *
> */
> extern void rotate_x(rvec x[], int nr, matrix rmat);
>
>
>
>
More information about the gromacs.org_gmx-developers
mailing list